Aggregate Mosquito Alert participation counts by Spanish municipality, 2018–2026

Aggregate Mosquito Alert participation counts for Spanish municipalities, 2018–2026, deposited as the input data for the IDAlert Task 5.4 randomized advertising experiment. All files are municipality-level counts: they contain no identifiers, no coordinates, no dates finer than a 60-day window, and no per-participant records. Masked background tracks. Participation is measured from background location tracks emitted passively by the Mosquito Alert app. Track locations are masked to a 0.025° × 0.025° grid (roughly 6 km² at Spanish latitudes) before any aggregation, so no precise location is used at any stage of this study or present in these files. Each masked cell is attributed whole to the municipality containing its centroid; because cells are coarse relative to municipalities, municipality boundaries are correspondingly approximate. Files and checksums File md5 Content participants_spain_municipality_aug_windows.csv ee47f0e841790fcba139aaecf892bb9d Distinct participants emitting masked background location tracks, by municipality × season × 60-day window, 8,244 municipalities, 2018–2026. Modal attribution (see below). participants_spain_province_aug_windows.csv ebd07219d817355eb60bd590af37e329 The same counts aggregated to province, deduplicated within province. municipality_reporter_counts.csv 7b6dbf8559d40899a57052992de3ecfd Distinct participants submitting mosquito reports, and report counts, by municipality × season × window, 2021–2025 (see “Coverage” below). Checksums The first file is byte-identical to the input recorded in manifest_2026_final.txt in the study repository. If you download this file, confirm the md5 above, and run analysis/r/03_randomization/assign_treatment_2026.R, the treatment assignment regenerates bit-for-bit (assignment_2026_final.csv, md5 cd9c86726fd7f9aa5c8e488d0b5361b2). These are pre-treatment data, deposited before the advertising campaign's outcome window opens. The 2026 after rows are placeholders (window_complete = FALSE, days_observed = 0, counts of 0), not outcomes. Outcome extracts will be added as a new version of this record after 14 October 2026. Attribution Participant counts use modal attribution. Each participant is counted in exactly one municipality: the one where they were observed on the greatest number of distinct days over their full track history — that is, where their sampling effort is concentrated. Ties are broken deterministically. This is not a claim about residence; the data cannot establish where anyone lives. Because each person is counted once, municipality counts sum to the province totals, which is how attribution can be verified from the files themselves (ratio 1.00 = an exact partition; presence attribution, in which a participant is counted in every municipality visited, gives 1.17–1.28). Note that the column is named n_participants for continuity with earlier releases even though it carries modal-attributed counts. Reporter counts use report-location attribution (point-in-polygon on the report coordinates). Reporting identifiers are deliberately not linkable to background-tracking identifiers, so modal attribution cannot be computed for reporters; most reporters submit a single report, so a modal rule would in any case collapse to the report location. Coverage The participant counts cover 2018–2026, including the complete 2026 pre-campaign window. The reporter counts cover 2021–2025: the report export underlying this version ends one day before the 2026 pre-window closes, so no 2026 season is emitted rather than a partial one. The 2026 reporter counts will be included in the next version of this record, alongside the outcome extracts. Windows Each season is split by a 15 August anchor into a pre window (16 June – 14 August) and a post window (16 August – 14 October). The anchor day belongs to neither window. window_complete and days_observed record whether the underlying data cover the whole window. Relation to other Mosquito Alert data Mosquito Alert publishes sampling-effort data at finer resolution (participants per masked 0.025° grid cell per day) and publishes all reports. The counts here are a coarser aggregation of the same underlying streams — the same masked cells, grouped into municipalities and 60-day windows — prepared for municipality-level analysis; the modal attribution is the one element not derivable from those releases, since they contain counts rather than per-participant records. Related resources Code, treatment assignment, and checksummed manifest: the study's GitHub repository (cite the tagged commit). Pre-registration: OSF registration [DOI to be added once the registration is public], which specifies the hypotheses, design, and analysis plan these data support. Licence: CC-BY-4.0. Part of IDAlert (Infectious Disease decision-support tools and Alert systems), which has received funding from the European Union's Horizon Europe programme under Grant Agreement 101057554.